Assessing fire effects on forest spatial structure using a fusion of Landsat and airborne LiDAR data in Yosemite National Park

Highlights

• Analyzed forest structure & fire severity using Landsat and LiDAR data at Yosemite

• Fire was a dominant process creating structural differences across burn severities.

• Fire resulted in loss of canopy in all height strata and retained multistory clumps.

• Forest types had individual structural trajectories with increasing fire severity.

• Low/moderate-severity fires best replicate historic forest structures.

Abstract

Mosaics of tree clumps and openings are characteristic of forests dominated by frequent, low- and moderate-severity fires. When restoring these fire-suppressed forests, managers often try to reproduce these structures to increase ecosystem resilience. We examined unburned and burned forest structures for 1937 0.81 ha sample areas in Yosemite National Park, USA. We estimated severity for fires from 1984 to 2010 using the Landsat-derived Relativized differenced Normalized Burn Ratio (RdNBR) and measured openings and canopy clumps in five height strata using airborne LiDAR data. Because our study area lacked concurrent field data, we identified methods to allow structural analysis using LiDAR data alone. We found three spatial structures, canopy-gap, clump-open, and open, that differed in spatial arrangement and proportion of canopy and openings. As fire severity increased, the total area in canopy decreased while the number of clumps increased, creating a patchwork of openings and multistory tree clumps. The presence of openings > 0.3 ha, an approximate minimum gap size needed to favor shade-intolerant pine regeneration, increased rapidly with loss of canopy area. The range and variation of structures for a given fire severity were specific to each forest type. Low- to moderate-severity fires best replicated the historic clump-opening patterns that were common in forests with frequent fire regimes. Our results suggest that managers consider the following goals for their forest restoration: 1) reduce total canopy cover by breaking up large contiguous areas into variable-sized tree clumps and scattered large individual trees; 2) create a range of opening sizes and shapes, including ~ 50% of the open area in gaps > 0.3 ha; 3) create multistory clumps in addition to single story clumps; 4) retain historic densities of large trees; and 5) vary treatments to include canopy-gap, clump-open, and open mosaics across project areas to mimic the range of patterns found for each forest type in our study.

Introduction

In frequent-fire pine and mixed-conifer forests in western North America (hereafter, dry forests), historic accounts (Dunning, 1923, Show and Kotok, 1924) and studies of forests with active fire regimes (Collins and Stephens, 2010, Collins et al., 2007, Larson and Churchill, 2012, Stephens and Collins, 2004, Stephens and Gill, 2005) have emphasized the importance of spatial variability in forest structure to maintain ecosystem process and resilience. A recent review of studies of stand-level structure found that fire-frequent dry forests were composed of mosaics of widely-spaced individual trees, tree clumps (two to 20 + trees), and openings (Larson & Churchill, 2012). Historically, these patterns of individual trees, tree clumps, and openings were maintained by fire and insect-driven mortality, and once established, tended towards self-perpetuation. Openings would act to moderate fire and inhibit bark-beetle dispersal (Finney et al., 2007, Pimont et al., 2011, Stephens et al., 2008) while the fine-scale local variation in canopy height and continuity would impede crown fires (Beaty and Taylor, 2007, Parisien et al., 2010, Pimont et al., 2011, Stephens et al., 2008, Thaxton and Platt, 2006). Openings also provided areas for subsequent regeneration, particularly of shade-intolerant, fire-resistant species, creating a fine-scale shifting mosaic maintained by frequent fire (Agee, 1993, Boyden et al., 2005, Cooper, 1960, Sánchez Meador et al., 2009).

Today, decades of fire exclusion have altered forest structure and often led to forests with nearly continuous canopies (Hessburg, Agee, & Franklin, 2005). Openings, especially large ones that can act as fire breaks and regeneration sites, are less prevalent than they were a century ago (Hessburg et al., 2005, Lutz et al., 2012, Scholl and Taylor, 2010). To restore structure, maintain resilience, and mitigate the possibility of large areas of high-severity fire, managers use mechanical thinning and prescribed and wildland fire across hundreds of thousands of hectares of public forests annually (Miller et al., 2012, North et al., 2012, Schoennagel and Nelson, 2011).

Researchers and managers need spatially-explicit measurements of tree clumps and openings over large areas to understand the ecological relationships between fire and the spatial structure of forests. Stem maps of reconstructed pre-Euro-American era forests or active-fire regime sites have been the primary source of information (e.g., Harrod, McRae, & Hartl, 1999). However, only 22 stem-map studies have been conducted on dry forest reference sites from 1960 to 2011 covering a cumulative 294.7 ha (Larson and Churchill, 2012, Lutz et al., 2012). The limited area suggests that the full diversity of spatial structures on western landscapes has been under sampled. Most spatially explicit tree maps are of small areas (0.5 to 4 ha) and thus do not inform managers on how pattern varies over spatial extents commonly used in restoration treatments (10 to 100 ha), or intact landscapes (> 1000 ha). In addition, few stem map studies contain height information, and little is known about the vertical structure of tree clumps. Silvicultural methods are being developed to restore stand-level patterns of tree clumps and openings (Churchill et al., 2013, North and Sherlock, 2012), but these lack high resolution spatial reference information over large scales (Larson & Churchill, 2012).

Airborne Light Detection and Ranging (LiDAR) data can assess forest structure over large areas (Hudak et al., 2009, Lefsky et al., 2002, Reutebuch et al., 2005) including patterns of gaps and tree clumps. LiDAR's strength is the high resolution (typically several measurements per square meter) and consistent measurement of ground elevation and canopy heights over large areas with greater fidelity to structural attributes than possible with satellite images (Asner et al., 2011, Hummel et al., 2011). Researchers have traditionally correlated LiDAR canopy measures with extensive ground-based tree measurements (e.g., for biomass or cubic volume). However, many forest LiDAR acquisitions lack concurrent field data. Lefsky, Hudak, Cohen, and Acker (2005) and Kane, McGaughey, et al. (2010) laid out the theoretical basis and provided a practical example (Kane, Bakker et al., 2010) for interpreting relative differences in forest structure using LiDAR data as a primary data source. Recently, researchers have begun to use LiDAR as a primary data source to study forest canopy structure without reference to field data over large areas (Asner et al., 2013, Kane et al., 2011, Kane et al., 2013, Kellner and Asner, 2009, Whitehurst et al., 2013). One of our goals is to identify methods to study openings and tree clumps for acquisitions that lack field data and demonstrate potential use for ecological analysis. Building on methods of Kane et al. (2011), we examine spatial structure of unburned stands and stands following fire. We used Landsat images to estimate fire severity across a 26 year period (1984 to 2010).

In this study, we use LiDAR data to examine the effects of different fire severities on the range of opening and tree clump structures (Fig. 1) found in three unburned and burned forest types (ponderosa pine, white fir-sugar pine, and red fir) common on the Sierra Nevada's western slope. While the role of fire in shaping and maintaining dry forests with active fire regimes is well documented (Collins and Stephens, 2010, Collins et al., 2007, Larson and Churchill, 2012, Stephens and Collins, 2004, Stephens and Gill, 2005), the effect of re-introduced fire following decades of fire exclusion is less well understood (but see Collins et al., 2011, Lydersen and North, 2012, Miller and Safford, 2012).

We used the methods identified for this study to address three questions related to the spatial structure of forests with increasing fire severity:

• How do the spatial structures of clumps and openings change with increasing fire severity for these three forest types?

• Which model(s) of forest restructuring (thin from below, dispersed mortality of all tree heights, or patchy mortality of all tree heights) best explains changes in structure with increasing fire severity?

• What are the management implications for forest structural restoration?